Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
MM: Fachverband Metall- und Materialphysik
MM 6: Computational Materials Modelling: Defects / Alloys
MM 6.4: Talk
Monday, September 5, 2022, 16:30–16:45, H44
Impurity segregation at grain boundaries in bcc iron: large scale models based on machine learned interatomic potentials — •Petr Šesták1, Monika Všianská1,2, Pavel Lejček1,3, and Miroslav Černý1 — 1Central European Institute of Technology, CEITEC BUT, Brno University of Technology, Purkyňova 123, CZ-616 69 Brno, Czech Republic — 2Department of Chemistry, Faculty of Science, Masaryk University, Kotlářská 2, CZ-611 37 Brno, Czech Republic — 3Institute of Physics of the Czech Academy of Sciences, Na Slovance 2, CZ-182 21 Prague 8, Czech Republic
In this work, we employed on the fly machine learning (ML) as it is implemented in the current version of the VASP to study segregation of Sn, P and Ge atoms at selected GBs (e.g. Σ3(112), Σ3(111), Σ5(310), Σ5(210), Σ13(510), Σ13(320), etc.) in bcc iron. Segregation energies were obtained using small supercells (<100 atoms), medium cells (∼400 atoms) and also larger cells (>1000 atoms). Data obtained from the small and medium cells were compared with results of ab initio calculations as well as with available experimental data and predictions. This comparison serves as a benchmark of the interatomic potentials received from ML. The comparison shows that the segregation energies and some other GBs characteristics obtained for small cells are very consistent with ab initio simulations which bring a proof of reliability of interatomic potentials. The obtained results revealed that using small simulation cells (leading to high concentrations of impurity atoms), typical for ab initio simulations, might not be sufficient to predict correct segregation energies.